Classifications

• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L13/00—Speech synthesis; Text to speech systems
  • G10L13/02—Methods for producing synthetic speech; Speech synthesisers
  • G10L13/027—Concept to speech synthesisers; Generation of natural phrases from machine-based concepts
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
  • G10L21/003—Changing voice quality, e.g. pitch or formants
  • G10L21/007—Changing voice quality, e.g. pitch or formants characterised by the process used
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/045—Combinations of networks
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
  • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
  • G10L21/0272—Voice signal separating
  • G10L21/0308—Voice signal separating characterised by the type of parameter measurement, e.g. correlation techniques, zero crossing techniques or predictive techniques
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
  • G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
  • G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/08—Learning methods
  • G06N3/088—Non-supervised learning, e.g. competitive learning
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L15/00—Speech recognition
  • G10L15/08—Speech classification or search
  • G10L15/16—Speech classification or search using artificial neural networks
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L17/00—Speaker identification or verification techniques
  • G10L17/18—Artificial neural networks; Connectionist approaches
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
  • G10L21/003—Changing voice quality, e.g. pitch or formants
  • G10L21/007—Changing voice quality, e.g. pitch or formants characterised by the process used
  • G10L21/013—Adapting to target pitch
  • G10L2021/0135—Voice conversion or morphing

Definitions

• This disclosure relates to unsupervised learning of disentangled speech content and style representation.
• Speech waveforms are a complex, high-dimensional form of data influenced by a number of underlying factors, which can be broadly categorized into linguistic contents and speaking styles.
• Learning disentangled latent representations from speech has a wide set of applications in generative tasks, including speech synthesis, data augmentation, voice transfer, and speech compression.
• One aspect of the disclosure provides a linguistic content and speaking style disentanglement model including a content encoder, a style encoder, and a decoder.
• the content encoder is configured to receive, as input, input speech, and generate, as output, a latent representation of linguistic content for the input speech.
• the content encoder trained to disentangle speaking style information from the latent representation of linguistic content.
• the style encoder is configured to receive, as input, the same or different input speech, and generate, as output, a latent representation of speaking style for the same or different input speech
• the style encoder is trained to disentangle linguistic content information from the latent representation of speaking style.
• the decoder is configured to generate output speech based on the latent representation of linguistic content for the input speech and the latent representation of speaking style for the same or different input speech.
• Implementations of the disclosure may include one or more of the following optional features
• the content encoder generates the latent representation of linguistic content as a discrete per-timestep latent representation of linguistic content that discards speaking style variations in the input speech.
• SUBSTITUTE SHEET (RULE 26) encoder may include: one or more convolutional layers configured to receive the input speech as input and generate an initial discrete per-timestep latent representation of the linguistic content: and a vector quantization (VQ) layer configured to apply an information bottleneck with straight-through gradients on each initial discrete per- timestep latent representation of the linguistic content to generate the latent representation of linguistic content as a sequence of latent variables representing the linguistic content form the input speech.
• the content encoder may be trained using a content VQ loss based on the latent representations of linguistic content generated for each timestep, whereby the VQ loss encourages the content encoder to minimize a distance between an output and a nearest codebook.
• the style encoder includes: one or more convolutional layers configured to receive the input speech as input; and a variational layer with Gaussian posterior configured to summarize an output from the one or more convolutional layers with a global average pooling operation across the time-axis to extract a global latent style variable that corresponds to the latent representation of speaking style.
• the global latent style variable may be sampled from a mean and variance of style latent variables predicted by the style encoder, and during inference, the global latent style variable may be sampled from the mean of the global latent style variables predicted by the style encoder.
• the style encoder may be trained using a style regularization loss based on a mean and variance of style latent variables predicted by the style encoder, wherein the style encoder uses the style regularization loss to minimize a Kullback-Leibler (KL) divergence between a Gaussian posterior with a unit Gaussian prior
• KL Kullback-Leibler
• the decoder is configured to: receive, as input, the latent representation of linguistic content for the input speech and the latent representation of speaking style for the same input speech; and generate, as output, the output speech comprising a reconstruction of the input speech.
• the model may be trained using a reconstruction loss between the input speech and the reconstruction of the input speech output from the decoder
• the decoder is configured to: receive, as input, the latent representation of linguistic content for the input speech and the latent representation of speaking style for the different input speech; and generate, as output, the output speech comprising linguistic content information specified by the input speech and speaking style information specified by the different input speech.
• the content encoder and the style encoder may be trained using a mutual information loss to minimize mutual information captured m the latent representations of linguistic content and speaking style.
• Another aspect of the disclosure provides a computer-implemented method when executed on data processing hardware causes the data processing hardware to perform operations that include receiving input speech and processing, using a content encoder, the input speech to generate a latent representation of linguistic content for the input speech.
• the content encoder is trained to disentangle speaking style information from the latent representation of linguistic content.
• the operations also include processing, using a style encoder, the same or different input speech to generate a latent representation of speaking style for the same or different input speech, and processing, using a decoder, the latent representation of linguistic content for the input speech and the latent representation of speaking style for the same or different input speech to generate output speech.
• the style encoder is trained to disentangle linguistic content information from the latent representation of speaking style.
• processing the input speech to generate the latent representation of linguistic content includes processing the input speech to generate the latent representation of linguistic content as a discrete per-timestep latent representation of linguistic content that discards speaking style variations in the input speech.
• the content encoder may include: one or more convolutional layers configured to receive the input speech as input and generate an initial discrete per-timestep latent representation of the linguistic content, and a vector quantization (VQ) layer configured to apply an information bottleneck with straight-through gradients on each initial discrete per- timestep latent representation of the linguistic content to generate the latent
• SUBSTITUTE SHEET (RULE 26) representations of linguistic content and speaking style from speech has a wide set of applications in generative tasks, including speech synthesis, data augmentation, voice transfer, and speech compression. Disentangling latent representations from speech can also be helpful for downstream tasks such as automated speech recognition and speaker classification.
• Implementations herein are directed toward unsupervised representation learning for speech by learning both global and localized representations. Specifically, implementations are directed toward disentangling latent representations of linguistic content and speaking style from speech using an autoencoder model that includes a content encoder, a style encoder, and a decoder
• the autoencoder model may be interchangeably referred to as a linguistic content and speaking style disentanglement model.
• the content encoder is associated with a local encoder having a vector-quantized ( VQ) layer configured to learn discrete per-timestep (e.g..
• the decoder reconstructs an input speech sample to the content and style encoders using the latent representations of linguistic content and speaking style output from the content and style encoders.
• the decoder may generate a new speech feature by combining the latent representation of linguistic content extracted from a first speech
• SUBSTITUTE SHEET (RULE 26) sample and the latent representation of speaking style extracted from a different second speech sample.
• a mutual information loss is applied to the content and style encoders to minimize mutual information from the latent representations of linguistic content and speaking style output from the content and style encoders.
• training the content and style encoders to minimize extracting mutual information in their respective outputs further disentangles the latent representations of linguistic content (e.g., local representations) and the latent representations of speaking style (e.g., global representations) from speech.
• the autoencoder model includes a content encoder, a style encoder 130, and a decoder.
• the decoder is configured to receive both content and style latent representations as input, and generate speech features as output That is, the decoder is configured to reconstruct the input speech as the output speech features.
• the trained autoencoder model may omit the use of the decoder
• FIG. 1 shows an example system 10 for training an autoencoder model (e.g , linguistic content and speaking style disentanglement model) 100 to disentangle latent representations of linguistic content 120 and speaking style 140 from input speech 102.
• the input speech 102 may include a sequence of speech features.
• the sequence of speech features representing the input speech 102 include log- MelTilterbank features.
• the system 10 includes a computing system 20 having data processing hardware 22 and memory hardware 24 in communication with the data processing hardware 22 and storing instructions that cause the data processing hardware 22 to perform operations
• the computing system 20 e g., the data processing hardware 22
• the training process aims to reconstruct each sample of input speech 102 by decoding latent representations of linguistic content 120 and speaking style 140 extracted from the input speech 102 and combined to generate corresponding speech features 152.
• the training process may store the latent representations of linguistic content 120 and speaking style 140 extracted from each sample of input speech
• the computing system 20 executes the trained autoencoder model 100 that includes a content encoder 110, a style encoder 130, and a decoder 150 to generate new speech features 152 as synthesized speech that conveys the linguistic content extracted from a first speech sample 50, 50a and having a speaking style extracted from a second speech sample 50, 50b.
• the trained autoencoder model 100 that includes a content encoder 110, a style encoder 130, and a decoder 150 to generate new speech features 152 as synthesized speech that conveys the linguistic content extracted from a first speech sample 50, 50a and having a speaking style extracted from a second speech sample 50, 50b.
• the content encoder 110 is configured to predict per-timestep latent representations of linguistic content 120 from the first speech sample 50a spoken by a first speaker while the style encoder 130 is associated with a variational autoencoder (VAE) configured to extract per-utterance latent representations of speaking style 140 from the different second speech sample 50b spoken by a different speaker.
• VAE variational autoencoder
• the first and second speech samples 50a, 50b spoken by the different speakers may include different linguistic content and the first and second speakers may speak with different speaking styles (e.g. prosody/accent).
• the trained autoencoder model 100 is adapted for use in a voice transfer application (e g...
• the trained autoencoder model 100 may similarly be applied in data augmentation applications where latent representations of speaking style 140 are extracted from speech samples spoken by different users and conveying different speaking styles to produce new speech features 152 from the decoder 150 that convey different augmentations of synthesized speech for a same linguistic content.
• SUBSTITUTE SHEET (RULE 26) identification model that is trained on the output speech features 152.
• the output 190 may not receive the speech features 152 and instead receive latent representations of linguistic content 120 and speaking style 140 from the content and style encoders 110, 130. respectively.
• FIG. 2 shows an example of the autoencoder model 100 for disentangling thelatent representations of linguistic content 120 and speaking style 140 from the input speech 102
• the autoencoder model includes the content encoder 110, the style encoder 130. and the decoder 150
• the decoder 150 is configured to receive both content and style latent representations 120, 140 as input, and generate speech features 152 as output.
• the trained autoencoder model 100 may omit the use of the decoder 150.
• the content encoder 1 10 may include a neural network having one or mors convolutional layers 1 12 and a vector quantization (VQ) layer 114.
• the content encoder 110 includes ten (10) 1 -dimensional convolutional layers with residual connections.
• a time stride of two may be applied in the third convolutional layer to reduce a final output length in half (e.g., from T to T/2).
• the one or more convolutional layers 112 of the content encoder 110 receives the input speech 102 to generate an initial latent representation 113 from the speech 102 and the VQ layer 114 applies an information bottleneck with straight-through gradients on the initial latent representation 113 to capture necessary localized information, such as phoneme or subword-like latent representations, and discard speaking style variations.
• the VQ layer 114 extracts a sequence of latent variables, represent the linguistic content from the input speech 102.
• the content encoder 110 is trained to learn to predict a discrete per-timestep (e.g., per frame) latent representation of linguistic content 120 (ct, .. ,CT) that discards, or is otherwise disentangled from, speaking style variations in the input speech 102.
• a VQ loss module 122 determines a content VQ loss 124 for the content encoder 110 based on the latent representations of linguistic content 120 predicted for each timestep.
• the content VQ loss 124 encourages
• the VQ loss module 122 may determine the content VQ loss 124 using the following equation. where 5g (•) denotes stop gradient operations.
• the content VQ loss 124 corresponds to a self-supervised loss similar to a clustering loss where the model is encouraged to cluster continuous data points and move other data points close to clusters.
• the style encoder 130 may include a neural network having one or more convolutional layers 132 and a variational layer 134 with Gaussian posterior configured to summarize the output from the convolution layer 132 with a global average pooling operation across the time-axis.
• the style encoder 130 processes the input speech 102 using six (6) residual one-dimensional convolutional layers with time stride two (2) on three different layers to result in 8 x time-length reduction.
• the global average pooling operation extracts a global latent style vanable, , that represents the speaking style 140 from the input speech 102.
• the global latent style variable s corresponding to the latent representation of speaking style 140 is sampled as follows where denotes a mean of the style variables predicted by the style encoder 130 and denotes a variance ot the style variables predicted by the style encoder. During inference, the global latent style variable s may simply correspond to the mean of the
• style variables predicted by the style encoder 130 from the input speech 102 Accordingly, the style encoder 130 is trained to learn to extract a per-utterance latent representation of speaking style 140 that is disentangled from linguistic content 120 in the input speech 102.
• a style loss module 142 may determine a style regularization loss 144, based on the mean and variance of the style variables predicted by
• the style encoder 130 uses the style regularization loss 144 to minimize a Kullback-Leibler (KL) divergence between a Gaussian posterior with a unit Gaussian prior
• the style loss module 142 may determine the style regularization loss 144 rising the following equation.
• the decoder 150 is configured to receive, as input, both the latent representations of linguistic content and speaking style 120, 140 output from the content and style encoders 110, 130, respectively, and reconstruct features 152, , during training During training, a reconstruction loss module 160 generates a reconstruction loss 162 between the input speech 102, fo serving as ground truth and the reconstructed speech 152, , for use in optimizing the model 100 to minimize both LI and L2-norm squared distance between Xand A”.
• the decoder 150 includes ten (10) 1 -dimensional convolution layers with residual connections fed speaking style information by concatenating the latent representation of speaking style 140, 5, in the channel axis on ⁇ 1, 3, 5, 7 ⁇ -th layers
• the quantized variables ci, .... er representing the linguistic content 120 from the input speech 102 may inherently capture non-content information such as speaking style information.
• the model 100 may estimate an minimize mutual information (MI) from the respective latent representations of linguistic content (eg ...,CT) 120 and speaking style (s) 140 output from the content and style encoders 110, 130.
• MI mutual information
• minimizing mutual information to reduce an amount of correlation between encoder outputs optimizes the model 100 to further disentangle the local and global representations of linguistic content and speaking style.
• a loss based on noise-contrastive estimation is used to estimate a lower bound MI between content and style as follows.
• the content encoder 110 receives input speech Xi and the style encoder 130 receives different input speech 102, X j , and the decoder 150 predicts speech features 152, to determine how well the speech features preserve the linguistic content of the original speech X
• an automated speech recognizer transcribed the predicted speech features and a word error rate is calculated for the transcription with ground-truth text for the original inpt.it speech X i fed to the content encoder 110.
• Table 1 below depicts the word error rates calculated for both the shuffle and non-shuffle scenarios with varying codebook sizes with and without mutual information loss
• SUBSTITUTE SHEET (RULE 26) [0042] .
• the trained model 100 may be used In a wide set of applications in generative tasks, including speech synthesis, data augmentation, voice transfer, and speech compression. The ability of the trained model 100 to disentangling these latent representations from speech can also be helpful for downstream tasks such as training automated speech recognition and/or speaker recognition/cl as sfo cation models.
• the trained model 100 can be used in two groups of applications: encoder-only applications and encoder-decoder applications Encoder-decoder applications, such as voice transfer applications and data augmentation applications, are discussed above with reference to FIG 1.
• FIGS. 4A shows a notable encoder-only application that includes using the trained style encoder 130 as a speaker recognition model 400 for speaker recognition tasks.
• the trained style encoder 130 is adapted for a speaker recognition application by pre-training the style encoder 130 on a large amount of uni abeled data including input speech corresponding to utterances spoken by one or more speakers Then, a small amount of labeled data fine-tunes the pre-trained style encoder to perform a few-shot speaker recognition task.
• the labeled data may vary the number of speech examples 402 per speaker, termed as “ 1 -shot” and “3 -shot”.
• a linear projection layer e.g.
• Softmax layer is overlain on top of the style encoder to compute logits for speaker classification based on latent representations of speaking style 140 output from the style encoder 130 for given input speech. During the fine-tuning, all parameters are frozen except for the projection layer.
• the style encoder 130 and projection layer 450 may form the speaker recognition model 400.
• the speaker recognition model 400a may be used for performing different speaker recognition tasks such as, without limitation, detecting different speakers (e.g., Di arization), computing speaker IDs, and/or voice matching for enrollment/verifi cation.
• FIG. 4B shows another encoder-only application that includes using the trained content encoder 110 as a speech recognition model 400b for generating speech recognition results for input speech 402.
• the content encoder 110 may be leveraged to extract latent representations of linguistic content 120 to provide local information for use
• SUBSTITUTE SHEET (RULE 26) (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), phase change memory (PCM) as well as disks or tapes.
• DRAM dynamic random access memory
• SRAM static random access memory
• PCM phase change memory
• the storage device 630 is capable of providing mass storage for toe computing device 600.
• the storage device 630 is a computer- readable medium.
• the storage device 630 may be a floppy disk device, a hard disk device, an optical disk device, or a tape device, a fl ash memory or other similar solid state memory device, or an array of devices, including devices in a storage area network or other configurations.
• a computer program product is tangibly embodied in an information earner.
• the computer program product contains instructions that, when executed, perform one or more methods, such as those described above.
• the information carrier is a computer- or machine-readable medium, such as toe memory 620, the storage device 630, or memory on processor 610.
• the high speed controller 640 manages bandwidth-intensive operations for the computing device 600, while the low speed controller 660 manages lower bandwidthintensive operations Such allocation of duties is exemplary only.
• the high-speed controller 640 is coupled to the memory 620, the display 680 (e.g., through a graphics processor or accelerator), and to the high-speed expansion ports 650, which may accept various expansion cards (not shown).
• the low-speed controller 660 is coupled to the storage device 630 arid a low-speed expansion port 690
• the low-speed expansion port 690 which may include various communication ports (e.g., USB, Bluetooth, Ethernet, wireless Ethernet), may be coupled to one or more input/output devices, such as a keyboard, a pointing device, a seamier, or a networking device such as a switch or router, e.g., through a network adapter
• the computing device 600 may be implemented in a number of different forms, as shown in the figure. For example, it may be implemented as a standard server 600a or multiple times in a group of such servers 600a, as a laptop computer 600b, or as part of a rack server system 600c.
• SUBSTITUTE SHEET performing instructions and one or more memory devices for storing instructions and data.
• a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e g., magnetic, magneto optical disks, or optical disks.
• mass storage devices for storing data, e g., magnetic, magneto optical disks, or optical disks.
• Computer readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including by way of example semiconductor memory devices, e.g , EPROM, EEPROM, and flash memory devices; magnetic disks, e.g , internal hard disks or removable disks; magneto optical disks; and CD ROM and DVD-ROM disks.
• the processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.
• one or more aspects of the disclosure can be implemented on a computer having a display device, e.g., a CRT (cathode ray tube), LCD (liquid crystal display) monitor, or touch screen for displaying information to the user and optionally a keyboard and a pointing device, e.g., a mouse or a trackball, by which the user can provide input to the computer.
• a display device e.g., a CRT (cathode ray tube), LCD (liquid crystal display) monitor, or touch screen for displaying information to the user and optionally a keyboard and a pointing device, e.g., a mouse or a trackball, by which the user can provide input to the computer.
• Other kinds of devices can be used to provide interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Health & Medical Sciences (AREA)
• Computational Linguistics (AREA)
• Multimedia (AREA)
• Audiology, Speech & Language Pathology (AREA)
• Human Computer Interaction (AREA)
• Acoustics & Sound (AREA)
• Theoretical Computer Science (AREA)
• Artificial Intelligence (AREA)
• Evolutionary Computation (AREA)
• Quality & Reliability (AREA)
• Signal Processing (AREA)
• Life Sciences & Earth Sciences (AREA)
• General Physics & Mathematics (AREA)
• Data Mining & Analysis (AREA)
• General Health & Medical Sciences (AREA)
• Molecular Biology (AREA)
• Computing Systems (AREA)
• General Engineering & Computer Science (AREA)
• Biophysics (AREA)
• Mathematical Physics (AREA)
• Software Systems (AREA)
• Biomedical Technology (AREA)
• Telephonic Communication Services (AREA)
• Machine Translation (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)

Priority Applications (4)

Applications Claiming Priority (2)

Publications (1)

Family

ID=78822351

Family Applications (1)

Country Status (6)

Families Citing this family (7)

Family Cites Families (4)

• 2021
  • 2021-11-18 EP EP21820421.2A patent/EP4244854A1/en not_active Withdrawn
  • 2021-11-18 US US17/455,667 patent/US20220189456A1/en active Pending
  • 2021-11-18 WO PCT/US2021/059991 patent/WO2022125290A1/en active Application Filing
  • 2021-11-18 JP JP2023535764A patent/JP2023553993A/ja active Pending
  • 2021-11-18 CN CN202180083495.7A patent/CN116635934A/zh active Pending
  • 2021-11-18 KR KR1020237022112A patent/KR20230116877A/ko unknown

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events